[PLUG] Debugging APM for 2.6 kernels - ideas?

[Elliott Mitchell]

>From: Keith Lofstrom <keithl at kl-ic.com>
> Unfortunately, like the 2.6 kernel that came with SUSE10, the pesky
> thing occasionally hangs coming out of suspend (blinking caps-lock
> and scroll-lock, often a sign of kernel panics).  I have not yet
> found a repeatable pattern, and I have tried a few dozen things to
> see if they stop the hangs.  I am using APM, having (as one of my
> frobs) turned off all the APIC stuff, since some folks on various
> lists out there think that is the problem.

I'm skeptical of this. Unless this is a really old machine (<800MHz) ACPI
is likely good. Or perhaps I'm biased against APM because on
multi-processor machines it is ACPI or nothing.

What this does sound like is borderline hardware/BIOS. If the BIOS isn't
careful and either assumes the machine is in a state that other OSes
never change or fail to restore such state suspend/resume become
Russian Roulette. Suspend/resume is always a bit of a risky proposition
as you're trusting the BIOS writers to of done their job well.


>From: "Jason R. Martin" <nsxfreddy at gmail.com>
> Since I'm guessing you can't see the kernel panic (since you mention
> the keyboard blinking) you probably need to figure out how to *get*
> the kernel panic output, since that would be the most useful to the
> developers.  If you are not seeing the dump because you're putting
> your laptop to sleep from X, then try putting it to sleep from a
> virtual console.  If that doesn't work, you'll have to try one of the
> crash dump patches or remote network crash dump patches.  I believe
> RedHat kernels have netdump built in, so you could investigate
> enabling that and setting up the netdump server on a remote machine. 
> Of course that might not work since going to S3 likely turns off your
> NIC, and I don't know how netdump works upon going back to S0.

Seems unlikely the network hardware would be brought back up if the rest
of the kernel fails on resume. Trying a VT does seem plausible as that is
more likely to come back on resume. Another optimistic approach if the
network hardware comes back would be to do remote-sysloging and see if
anything shows up there. A serial console would also be a place to look.


-- 
(\___(\___(\______          --=> 8-) EHM <=--          ______/)___/)___/)
 \BS (    |         EHeM at gremlin.m5p.com PGP 8881EF59         |    )   /
  \_CS\   |  _____  -O #include <stddisclaimer.h> O-   _____  |   /  _/
    \___\_|_/82 04 A1 3C C7 B1 37 2A*E3 6E 84 DA 97 4C 40 E6\_|_/___/